1Settimana
It's amazing that the post provoked so little reaction. đ€
The topic is very well presented and the instructions are very practical. And actually, quite a few people here claim to be practicing value investing.
I do have some questions about the value factor, but I don't know if anyone is interested. đ€·
The topic is very well presented and the instructions are very practical. And actually, quite a few people here claim to be practicing value investing.
I do have some questions about the value factor, but I don't know if anyone is interested. đ€·
âą
88
âą
1Settimana
@Epi the text is very long and only for people who are interested.
Feel free to ask your question, I would love to discuss it. Actually, everything is based exclusively on published research papers and can therefore be read and replicated. You could say that my text is merely a summary of the existing literature.
Feel free to ask your question, I would love to discuss it. Actually, everything is based exclusively on published research papers and can therefore be read and replicated. You could say that my text is merely a summary of the existing literature.
âą
33
âą1Settimana
@QMom I like summaries like that. But for most people it's probably too long and too abstract. So a little shorter and more concrete helps the thoughts to have more reach.
My question relates to the thesis that value is a factor in its own right. Can't the excess return be fully explained by the risk premium of certain stocks that appear to be cheap? In other words, if you discount the increased risk of value, then all that remains of the risk-adjusted premium is random noise.
I can't remember where, but I once read this in a paper on momentum with the thesis that there is no risk-adjusted premium apart from momentum. đ€·
My question relates to the thesis that value is a factor in its own right. Can't the excess return be fully explained by the risk premium of certain stocks that appear to be cheap? In other words, if you discount the increased risk of value, then all that remains of the risk-adjusted premium is random noise.
I can't remember where, but I once read this in a paper on momentum with the thesis that there is no risk-adjusted premium apart from momentum. đ€·
âą
22
âą
1Settimana
@Epi
Couldn't we jump straight to Carhart (1997)?
Couldn't we jump straight to Carhart (1997)?
âą
22
âą1Settimana
@Epi This is a very good point, because one of the theses in Fama and French (2015) is exactly this. They argue that with the addition of CMA value would have become redundant.
However, this seems to be clearly dependent on design.
In statistics, we can perform a spanning regression test to test this effect. Such a test was performed for the enhanced value factor as described above by Hanauer & Blitz (2021). The slightly enhanced factor provides significant alpha, regardless of whether HML is included in the regression or not. It also shows a clear value tendency, but is not identical to the classic HML factor and thus contains additional information. The strong negative correlation with momentum is also striking, which confirms the well-known trade-off between value and momentum. At the same time, the positive exposures to profitability and conservative investment policy show overlaps with quality characteristics, which indicates that the strategy also captures something of quality in addition to classic value. However, it is not redundant and is not explained by the standard models.
(Unfortunately I cannot attach the table)
Ultimately, however, this is a very good point: each factor should show independent returns in the long/short format. Since there are now well over 100 factors, this topic is covered in Swade et al. (2023) Factor Zoo (.zip) (highly recommended).
However, this seems to be clearly dependent on design.
In statistics, we can perform a spanning regression test to test this effect. Such a test was performed for the enhanced value factor as described above by Hanauer & Blitz (2021). The slightly enhanced factor provides significant alpha, regardless of whether HML is included in the regression or not. It also shows a clear value tendency, but is not identical to the classic HML factor and thus contains additional information. The strong negative correlation with momentum is also striking, which confirms the well-known trade-off between value and momentum. At the same time, the positive exposures to profitability and conservative investment policy show overlaps with quality characteristics, which indicates that the strategy also captures something of quality in addition to classic value. However, it is not redundant and is not explained by the standard models.
(Unfortunately I cannot attach the table)
Ultimately, however, this is a very good point: each factor should show independent returns in the long/short format. Since there are now well over 100 factors, this topic is covered in Swade et al. (2023) Factor Zoo (.zip) (highly recommended).
âą
22
âą1Settimana
@Savvy_investor_2000 The Carhart model first added only momentum, which is negatively correlated with value. Value therefore only became redundant with the addition of quality as an original HML factor.
âą
22
âą1Settimana
@QMom
Hardly anyone drives it in isolation anymore, at most embedded in multi-factor combos. It still has its place there.
Hardly anyone drives it in isolation anymore, at most embedded in multi-factor combos. It still has its place there.
âą
33
âą1Settimana
@Savvy_investor_2000 I agree with you 100%. Value should always be in proportion.
âą
33
âą
@Epi factors are generally referred to as the risk premium. Therefore, the risk-adjusted return does not necessarily have to be different from that of the general market (beta), for example.
You would only have a correlation <1.
Value would be a factor insofar as this factor is one of the factors used by Fama&French and others to explain market behavior.
As far as I know, momentum provides the highest Sharpe ratio. Momentum is not included in the Fama & French model. Some portfolio managers use momentum in integrated models, where momentum is included in the selection of stocks but does not dominate them. This is an attempt to avoid high trading costs.
Value would definitely be a factor for me, because as far as I know the deviation from beta is already systematic. If I don't achieve a risk-adjusted return, I can still try to use the reduced correlation. For me, for example, the combination of value and momentum is very attractive.
Either the people here are using ChatGPT to pepper the stuff around their ears or they are absurdly deep into the subject đ because what surprises me is the intense jump between Fama&French and Carhart.
@QMom You say quality combined with value? I'm thinking at the individual value level then? At the factor level, I seem to remember that value tends to correlate negatively with profitability, which could lead to problems if I mix quality and value, as I would somewhat destroy the value effect. On an individual stock level it could work by looking for stocks that are positive for value and quality.
Therefore, because I am in ETFs, I am considering killing the $XDEQ and only going for momentum, value and SC value. Since momentum and SC value are highly procyclical and value is at least still positively correlated, you have to be able to withstand something.
You would only have a correlation <1.
Value would be a factor insofar as this factor is one of the factors used by Fama&French and others to explain market behavior.
As far as I know, momentum provides the highest Sharpe ratio. Momentum is not included in the Fama & French model. Some portfolio managers use momentum in integrated models, where momentum is included in the selection of stocks but does not dominate them. This is an attempt to avoid high trading costs.
Value would definitely be a factor for me, because as far as I know the deviation from beta is already systematic. If I don't achieve a risk-adjusted return, I can still try to use the reduced correlation. For me, for example, the combination of value and momentum is very attractive.
Either the people here are using ChatGPT to pepper the stuff around their ears or they are absurdly deep into the subject đ because what surprises me is the intense jump between Fama&French and Carhart.
@QMom You say quality combined with value? I'm thinking at the individual value level then? At the factor level, I seem to remember that value tends to correlate negatively with profitability, which could lead to problems if I mix quality and value, as I would somewhat destroy the value effect. On an individual stock level it could work by looking for stocks that are positive for value and quality.
Therefore, because I am in ETFs, I am considering killing the $XDEQ and only going for momentum, value and SC value. Since momentum and SC value are highly procyclical and value is at least still positively correlated, you have to be able to withstand something.
âą
11
âą@SchlaubiSchlumpf Thank you for your helpful comment. I think you have to make a fundamental distinction between theory and practical application and between long/short and long-only implementation.
You have already described this in your answer to Val/Mom. From a theoretical point of view, the combination of value and quality generally makes sense, as both factors correlate negatively (L/S). At portfolio level, this means that they contain different information and a combination leads to diversification effects. Ultimately, both should generate alpha and have positive expected returns.
At the construction level, it can be seen that value is often more effective in relation to a quality multiple than a pure value signal. Piotroski (2000) combines value with quality characteristics and thus filters out weak companies. A similar effect can already be seen in the choice of multiple within the value factor. If, for example, P/S is compared with EV/EBITDA, EV/EBITDA shows a better performance. The explanation is simple: EV/EBITDA already contains qualitative information about the profitability and capital structure of the company, whereas pure price multiples contain little information.
In order to decide in practice whether a company is actually "cheap", the ratio of quality to price is more useful than a price figure alone. A "good company at a low price" is clearly more attractive than a "bad company at a low price". In both cases, we expect a multiple expansion in the direction of fair value, except that the fair level is higher for better quality companies.
Whether this is implemented as a cross-section between value and quality stocks, a combined factor is constructed or the signals are linked in some other way is not so decisive from a methodological point of view in my opinion.
You have already described this in your answer to Val/Mom. From a theoretical point of view, the combination of value and quality generally makes sense, as both factors correlate negatively (L/S). At portfolio level, this means that they contain different information and a combination leads to diversification effects. Ultimately, both should generate alpha and have positive expected returns.
At the construction level, it can be seen that value is often more effective in relation to a quality multiple than a pure value signal. Piotroski (2000) combines value with quality characteristics and thus filters out weak companies. A similar effect can already be seen in the choice of multiple within the value factor. If, for example, P/S is compared with EV/EBITDA, EV/EBITDA shows a better performance. The explanation is simple: EV/EBITDA already contains qualitative information about the profitability and capital structure of the company, whereas pure price multiples contain little information.
In order to decide in practice whether a company is actually "cheap", the ratio of quality to price is more useful than a price figure alone. A "good company at a low price" is clearly more attractive than a "bad company at a low price". In both cases, we expect a multiple expansion in the direction of fair value, except that the fair level is higher for better quality companies.
Whether this is implemented as a cross-section between value and quality stocks, a combined factor is constructed or the signals are linked in some other way is not so decisive from a methodological point of view in my opinion.
âą
11
âą@SchlaubiSchlumpf Examples of managers pursuing this thesis in the ETF sector are Dimensional and Avantis.
âą
11
âą